Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.misc > #4250 > unrolled thread

Google's email scanning

Back to article view | Back to comp.misc

Contents

  Google's email scanning Hils <hils@saynotospam.net> - 2014-08-05 05:44 +0100
    Re: Google's email scanning Mike Easter <MikeE@ster.invalid> - 2014-08-05 14:43 -0700
      Re: Google's email scanning Rich <rich@example.invalid> - 2014-08-06 01:34 +0000
        Re: Google's email scanning Mike Easter <MikeE@ster.invalid> - 2014-08-07 10:20 -0700
          Re: Google's email scanning Rich <rich@example.invalid> - 2014-08-08 23:12 +0000
            Re: Google's email scanning Mike Easter <MikeE@ster.invalid> - 2014-08-08 22:19 -0700
    Re: Google's email scanning john <nospam@no.net> - 2014-08-19 12:57 -0800
      Re: Google's email scanning Huge <Huge@nowhere.much.invalid> - 2014-08-20 08:45 +0000

#4250 — Google's email scanning

"Google has revealed the identity of a user storing images of child
abuse on his Gmail account to police in the US, flagging up the fact
that the US tech giant is scanning users email for illegal activity.

[...] This is the first time that the company has been shown to scan
emails for images of child abuse.

The company has previously said that it watches for this type of content
by creating a database of ‘hashes’ – unique codes that correspond to
known images of abuse and that allow the company to ‘look’ for abusive
content without examining the image itself or keeping a permanent
database of abusive content.

[...] The Independent understands that the company does not keep an eye
out for “other forms” of illegal activity such as digital piracy or hate
speech."

http://www.independent.co.uk/life-style/gadgets-and-tech/google-tips-off-us-police-to-man-storing-images-of-child-abuse-on-his-gmail-account-9647551.html

[toc] | [next] | [standalone]

#4252

Hils wrote:
> The company has previously said that it watches for this type of content
> by creating a database of ‘hashes’ – unique codes that correspond to
> known images of abuse and that allow the company to ‘look’ for abusive
> content without examining the image itself or keeping a permanent
> database of abusive content.

So, the child porn fiends who acquire or distribute such images should 
make some small change in the image so that the hash is different; if 
I'm understanding this policing mechanism.

There is probably some AI recognition tech afoot which could recognize 
(or at least suspect) child porn by some logic a little fuzzier than the 
precision of a hash on a specific and exact image subject to change.

-- 
Mike Easter

[toc] | [prev] | [next] | [standalone]

#4254

Mike Easter <MikeE@ster.invalid> wrote:
> Hils wrote:
> > The company has previously said that it watches for this type of
> > content by creating a database of ?hashes? ? unique codes that
> > correspond to known images of abuse and that allow the company to
> > ?look? for abusive content without examining the image itself or
> > keeping a permanent database of abusive content.

> So, the child porn fiends who acquire or distribute such images
> should make some small change in the image so that the hash is
> different; if I'm understanding this policing mechanism.

It is unlikely they are using cryptographic hashes, for just this very
reason.  They are probably using something similar to phash
(http://www.phash.org/) where small changes to the image produce close
enough hashes to tag "similar" to them.

[toc] | [prev] | [next] | [standalone]

#4272

Rich wrote:
> Mike Easter wrote:
>> Hils wrote:

>>> ca database of ?hashes? ? unique codes that
>>> correspond to known images of abuse and that allow the company to
>>> ?look? for abusive content without examining the image itself or
>>> keeping a permanent database of abusive content.

So the term 'unique' is not appropriate in this instance.

>> So, the child porn fiends who acquire or distribute such images
>> should make some small change in the image so that the hash is
>> different; if I'm understanding this policing mechanism.
>
> It is unlikely they are using cryptographic hashes, for just this very
> reason.  They are probably using something similar to phash
> (http://www.phash.org/) where small changes to the image produce close
> enough hashes to tag "similar" to them.

I don't really understand yet what/how phashing - perceptual hashing - 
works in the way I understand crypto hashing.  Wikipedia doesn't have an 
article.  It has one on digital video fingerprinting, which must be 
based on similar principles and algos.

It would seem to me that if a coder had a good understanding of how 
phashing or digital video (or audio) fingerprinting works, they/one 
could develop a strategy for making the least modifications 
algorithmically which would most adversely affect/neutralize the 
fingerprinting algos.



-- 
Mike Easter

[toc] | [prev] | [next] | [standalone]

#4276

Mike Easter <MikeE@ster.invalid> wrote:
> Rich wrote:
> > Mike Easter wrote:
> >> Hils wrote:

> >>> ca database of ?hashes? ? unique codes that
> >>> correspond to known images of abuse and that allow the company to
> >>> ?look? for abusive content without examining the image itself or
> >>> keeping a permanent database of abusive content.

> So the term 'unique' is not appropriate in this instance.

> >> So, the child porn fiends who acquire or distribute such images
> >> should make some small change in the image so that the hash is
> >> different; if I'm understanding this policing mechanism.
> >
> > It is unlikely they are using cryptographic hashes, for just this very
> > reason.  They are probably using something similar to phash
> > (http://www.phash.org/) where small changes to the image produce close
> > enough hashes to tag "similar" to them.

> I don't really understand yet what/how phashing - perceptual hashing - 
> works in the way I understand crypto hashing.  Wikipedia doesn't have an 
> article.  It has one on digital video fingerprinting, which must be 
> based on similar principles and algos.

> It would seem to me that if a coder had a good understanding of how 
> phashing or digital video (or audio) fingerprinting works, they/one 
> could develop a strategy for making the least modifications 
> algorithmically which would most adversely affect/neutralize the 
> fingerprinting algos.

Perceptual Hashing 
http://bertolami.com/index.php?engine=blog&content=posts&detail=perceptual-hashing

No idea on how good it is, just found the reference.

[toc] | [prev] | [next] | [standalone]

#4277

Rich wrote:
> Mike Easter
>> I don't really understand yet what/how phashing - perceptual hashing -
>> works in the way I understand crypto hashing.  Wikipedia doesn't have an
>> article.

> Perceptual Hashing
> http://bertolami.com/index.php?engine=blog&content=posts&detail=perceptual-hashing
>
> No idea on how good it is, just found the reference.

I like that one better than anything else I've found so far.  Tnx.


-- 
Mike Easter

[toc] | [prev] | [next] | [standalone]

#4494

Hils wrote:

> 
> [...] The Independent understands that the company does not keep an eye
> out for “other forms” of illegal activity such as digital piracy or hate
> speech."
> 

What is illegal hate speech? Is that a co.uk thing?

[toc] | [prev] | [next] | [standalone]

#4505

On 2014-08-19, john <nospam@no.net> wrote:
> Hils wrote:
>
>> 
>> [...] The Independent understands that the company does not keep an eye
>> out for “other forms” of illegal activity such as digital piracy or hate
>> speech."
>> 
>
> What is illegal hate speech? Is that a co.uk thing?

http://en.wikipedia.org/wiki/Hate_speech#United_Kingdom


-- 
Today is Boomtime, the 13th day of Bureaucracy in the YOLD 3180
We are all living in Israel; it's just that some of you haven't realised yet.

[toc] | [prev] | [standalone]

Back to top | Article view | comp.misc

csiph-web